Given a dataset of corporate entities, write a PySpark routine to deduplicate records based on fuzzy matching of names and registration numbers.
